Geography, Race, and Health

Summary of working paper 9513
Featured in print Bulletin on Aging & Health

The existence of racial disparities in medical treatment and health outcomes is well known. A recent comprehensive study by the Institute of Medicine concludes that "only a handful of the hundreds of studies reviewed here... find no racial and ethnic differences in care." (1) The study notes that racial differences could result from differences in access to care, socioeconomic or geographic factors, or racial differences in preferences of patients or attitudes of providers, but that more work is needed to determine the relative contribution of each factor.

In "Geography and Racial Health Disparities" (NBER Working Paper No. 9513), Amitabh Chandra and Jonathan Skinner argue that the influence of geography should be taken more seriously in measuring racial disparities and designing reforms to reduce them. The authors make several points to reach this conclusion.

First, they note that there are large geographic disparities in treatment and outcomes. For example, in 1996 there was an average of 7.5 angioplasty procedures performed per 1000 Medicare enrollees in the US, but the rate varied from 2.6 to 22.3 per 1000 across the 306 "hospital referral regions" (HRR) in the country. The authors find that substantial differences persist even after adjusting for differences in patient characteristics, and that large disparities exist not only across regions of the country but also within states and even cities.

Second, the authors find that people of different racial groups tend to seek care from different hospitals and physicians. The authors analyze admission patterns of Medicare patients treated for a heart attack and find that 50% of black patients are admitted to hospitals that combined account for just 14% of non-black admissions, not the 50% one would find if black and non-black patients were equally likely to go to each hospital. This may reflect residential segregation by race, but could also reflect the selection of different hospitals by blacks and non-blacks living in the same area.

Third, the authors note that racial disparities exhibit substantial geographic variation. For example, black and non-black Medicare patients in Massachusetts are equally likely to receive angioplasty, while blacks in Arkansas are only one-fourth as likely to receive angioplasty as similar non-blacks.

Building on these points, the authors demonstrate that ignoring region can produce a misleading estimate of racial disparities. For example, average expenditures for Hispanic and non-Hispanic Medicare beneficiaries are the same, suggesting no disparity. However, most Hispanic beneficiaries live in states where Medicare expenditures are higher; incorporating these regional differences, Hispanics may well have lower expenditures than non-Hispanics. In general, estimates of racial disparities may be too high or too low when region is ignored, as differences that are the consequence of where people live are mistakenly classified as being due to race.

The authors conclude that the policy implications of racial disparities depend on whether disparities result from geographic variation in treatment norms or from differential treatment by race within a hospital or physician's practice. If minorities tend to live in areas where quality of care is lower for all patients, then reducing geographic disparities in quality of care could significantly reduce racial disparities in treatment and health outcomes.

1 Smedley, Brian D. et. al. "Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington DC: National Academy Press, 2002.